Measuring surface temperatures in a woodland savanna: Opportunities and challenges of thermal imaging in an open-canopy ecosystem

https://doi.org/10.1016/j.agrformet.2021.108484Get rights and content

Abstract

In open-canopy ecosystems, thermal imaging affords an unprecedented opportunity to resolve concurrent temperatures of overstory vegetation, understory vegetation, and soil across space and time. This simultaneous view of ecosystem components promises a holistic understanding of ecosystem energy status, defines diverse thermal niches, and can provide a full suite of thermal measurements to drive ecosystem energy budget models. However, thermal imaging in open-canopy ecosystems also presents challenges: emissivity and background radiation data required for image calibration are variable across the scene; mixed pixels can be misleading because of divergent component temperatures; and targets of interest have very different pixel dimensions associated with their different distances from the camera. In this study, we evaluated effects of these challenges on calculated target temperatures, and we contextualized those results with five months of half-hourly thermal images, over vs. understory radiation measurements, ground-based emissivity estimates, and an application of thermal images to drive the two-source energy balance model (TSEB) in a Californian woodland savanna. We found that, though background radiation conditions varied considerably at different locations within the ecosystem, the high emissivities of the ecosystem components minimized the effect of that variation on calibrated target temperatures. Different pixel dimensions (i.e. variable geographical space covered by a single thermal image pixel) were associated with changes of temperature minima and maxima by over one degree Celsius, but they had little effect on aggregate summary values (e.g. estimates of mean temperature). Conversely, mixed pixels, given the relatively widely divergent component temperatures in our heterogeneous system, had the potential to influence calibrated target temperatures dramatically, by several degrees Celsius. The TSEB results corroborate these findings: they are sensitive to differences in component temperatures, while emissivity and reflected radiation corrections result in a negligible difference in sensible heat flux predictions.

Introduction

Plant temperature exerts fundamental control on physiological processes such as photosynthesis (Farquhar, von Caemmerer, Berry, 1980, Way and Yamori, 2014), respiration (Heskel et al., 2016), transpiration (Gates, 1968), and growth and development (Michaletz, 2018). It has long been recognized as an indicator of plant water relations (Brown and Escombe, 1905), and has been used to assess moisture stress at scales from individuals (Jackson et al., 1977) to continents (Anderson et al., 2007). At an ecosystem-scale, plant and soil temperatures influence fluxes of carbon, water, and energy and play a key role in energy budget closure (Heusinkveld, Jacobs, Holtslag, Berkowicz, 2004, Meyers, Hollinger, 2004).

Ecosystem structure can strongly influence vegetation thermal environments and therefore productivity (Rotenberg and Yakir, 2010). Open-canopy, semi-arid and Mediterranean ecosystems are subject to seasonal water deficit, fire, grazing pressure, and high incoming radiation, yet they are typically carbon sinks, even during extended drought (Ma et al., 2016, Rotenberg, Yakir, 2010). Structure (in addition to phenology [Maseyk et al., 2008], plant physiology [Baldocchi et al., 2004], and rooting depth [Miller et al., 2010]) appears to play an important role in a savanna’s ability to maintain function: the low density of trees results in high aerodynamic roughness and strong canopy-atmosphere coupling, which makes the system an efficient convector of sensible heat. Therefore, despite the high radiation loading, low albedo (compared to shrublands and grasslands), and seasonally very low precipitation, savannas can maintain a relatively low canopy surface temperature and physiologically favorable carbon, water, and energy balances (Baldocchi, Xu, Kiang, 2004, Rotenberg, Yakir, 2010). In this context, long-term and spatially resolved temperatures of the various components of a savanna system (e.g. soil, understory grass, and overstory trees) are of great interest.

Field-deployable thermal cameras are a promising tool with which to measure these temperatures (Aubrecht, Helliker, Goulden, Roberts, Still, Richardson, 2016, Kim et al., 2016, Kim, Still, Roberts, Goulden, 2018, Pau, Detto, Kim, Still, 2018, Still, Powell, Aubrecht, Kim, Helliker, Roberts, Richardson, Goulden, 2019). Aubrecht et al. (2016) provide an excellent summary of thermal camera calibration theory, a sensitivity analysis for calibration parameters, and a camera sensor noise and accuracy assessment. They also demonstrate the utility of thermal cameras to measure canopy temperature in an eastern deciduous forest and an evergreen needleleaf forest (Aubrecht et al., 2016). Here, we extend Aubrecht et al.’s thermal imaging guide to the case of heterogeneous, open-canopy ecosystems. While thermal imaging of an open-canopy ecosystem presents an opportunity to measure vertically and horizontally disparate components of the system concurrently, it may also introduce challenges which are (comparatively) absent when imaging a closed canopy. Specifically, background radiation conditions (i.e. radiation from the surroundings that reflects off the target of interest) vary significantly for regions of interest at different vertical heights, emissivity is variable across ecosystem components, thermal image pixels correspond to variable geographical space associated with the varying distances of targets from the camera, and ecosystem heterogeneity amplifies the confounding effects of mixed pixels.

In this paper, we: (1) quantify the theoretical effects of the potential challenges associated with thermal imaging in open-canopy systems; (2) assess the functional importance of these challenges in a Californian blue oak (Quercus douglasii) woodland savanna; and (3) contextualize the calibration challenges and demonstrate an opportunity offered by thermal imaging in an open-canopy system by using thermal images to generate sensible heat flux estimations, including geographical heterogeneity, with the two-source energy balance model (TSEB, Norman et al., 1995). Additionally, we include an analysis of the effect of the thermal camera’s protective enclosure on calibrated temperatures.

Section snippets

Site description

The camera deployment site is a seasonally-grazed oak savanna at 177 m of elevation in the lower foothills of California’s Sierra Nevada Mountains (38.438N, 120.968W). The overstory is open (leaf area index approximately 0.7, with considerable variation) and comprised mainly of deciduous Quercus douglasii (“oak”) interspersed with less numerous but considerably taller Pinus sabiniana (“pine”, Baldocchi et al., 2010, Fig. 1). The understory is dominated by annual C3 grass species, mostly

Imaging the reference panel

Each of the parameters required to convert the FLIR’s raw signal into temperature (Eqns. 2 and 3) was reported, measured, or calculated, with the exception of ϵsky, which was assumed to be unity (see Appendix A for detail); there was no statistical tuning. There were, however, calibration choices associated with the source of the reflected energy, the location of the temperature and relative humidity measurements, and the presence or absence of an enclosure window. Given the physical realities,

Conclusions

In our open-canopy savanna system, the effects of large temperature differences among ecosystem components dominated the effects of emissivity and reflected radiation diversity, both in calibration (mixed pixels are most confounding) and in application (choice of substrate has a larger effect on TSEB results than emissivity/reflected radiation corrections). This result highlights the value of spatially-explicit thermal measurements, which is further emphasized by the utility of multiple

Funding

This work was supported by NASA Earth and Space Science Fellowship award NNX16AO21H to M. Johnston and P. Moorcroft, the Harvard University Department of Organismic and Evolutionary Biology, and the European Union’s Horizon 2020 Research and Innovation program, Marie Sklodowska-Curie grant agreement No. 703978 to A. Andreu.

Declaration of Competing Interest

None.

Acknowledgements

The authors would like to thank Elsa Abbott for spectrometer measurements of the panel and the soil, Jonathan Franklin for suggesting a Sobel filter/FFT detection of FOV shifts, Eva Rubio for consultation on the emissivity box, Alfonso Torres-Rua for emissivity discussions, Hector Nieto for pyTSEb code and support, and Koen Hufkens, Don Aubrecht, Youngil Kim, Susan Meerdink, Tony Rockwell, and Steve Wofsy for instrumentation consultation. In addition, many thanks to two anonymous reviewers for

References (44)

  • F. Li et al.

    Utility of remote sensing-based two-source energy balance model under low- and high-vegetation cover conditions

    J Hydrometeorol

    (2005)
  • S. Ma et al.

    Inter-annual variability in carbon dioxide exchange of an oak/grass savanna and open grassland in California

    Agric For Meteorol

    (2007)
  • K.S. Maseyk et al.

    Physiology-phenology interactions in a productive semi-arid pine forest

    New Phytol

    (2008)
  • S. Michaletz

    Evaluating the kinetic basis of plant growth from organs to ecosystems

    New Phytol

    (2018)
  • S. Pau et al.

    Tropical forest temperature thresholds for gross primary productivity

    Ecosphere

    (2018)
  • E. Rubio et al.

    Thermal-infrared emissivities of natural surfaces: improvements on the experimental set-up and new measurements

    Int J Remote Sens

    (2003)
  • L. Song et al.

    Estimation of surface heat fluxes using multi-angular observations of radiative surface temperature

    Remote Sens Environ

    (2020)
  • L. Song et al.

    Application of remote sensing-based two-source energy balance model for mapping field surface fluxes with composite and component surface temperatures

    Agric For Meteorol

    (2016)
  • C. Still et al.

    Thermal imaging in plant and ecosystem ecology: applications and challenges

    Ecosphere

    (2019)
  • M. Anderson et al.

    A climatological study of evapotranspiration and moisture stress across the continental United States based on thermal remote sensing: 1. model formulation

    J Geophys Res

    (2007)
  • A. Andreu et al.

    Modeling surface energy fluxes over a dehesa (oak savanna) ecosystem using a thermal based two source energy balance model (TSEB) II -- integration of remote sensing medium and low spatial resolution satellite images

    Remote Sens

    (2018)
  • A. Andreu et al.

    Influence of component temperature derivation from dual angle thermal infrared observations on TSEB flux estimates over an irrigated vineyard

    Acta Geophys

    (2015)
  • Cited by (9)

    • Quantifying canopy conductance in a pine forest during drought from combined sap flow and canopy surface temperature measurements

      2022, Agricultural and Forest Meteorology
      Citation Excerpt :

      However, we found that the thermal infrared temperature slightly underestimated the temperature of the target reference, an observation also made by Kim et al. (2018). A recent study found instead that thermal infrared sensor overestimated leaf temperature under high solar radiation with differences being over 2 °C Johnston et al. (2021). In our case, we hypothesise that the small thermal inertia material used to built the target led to a high temporal variance of the target surface temperature.

    • What lies beneath: Vertical temperature heterogeneity in a Mediterranean woodland savanna

      2022, Remote Sensing of Environment
      Citation Excerpt :

      In addition to “bulk” canopy ROIs, as outlined in Fig. 3, we separately considered canopy tops and bottoms, as defined by the top and bottom thirds of ROI pixel rows. We followed methods detailed in Aubrecht et al. (2016) and Johnston et al. (2021) to correct the raw images for target emissivity, distance between sensor and target, air temperature, air transmissivity, reflected radiation (assumed originating from the sky), and the effects of the enclosure window (please see Table S2 for further details on instrumentation). We calculated transmissivity (0.971) and reflectivity (0.004) of the germanium protective enclosure window (Edmund Optics 8–12 μm anti-reflection coated) using curves provided by Edmund Optics, and we measured window temperature on the internal window surface with a thermocouple.

    • Effects of light and shadow on soil moisture content monitored by UAV thermal infrared remote sensing

      2023, Nongye Gongcheng Xuebao/Transactions of the Chinese Society of Agricultural Engineering
    View all citing articles on Scopus
    View full text